How hitagiiii/thesis-recommendation-system works
This is a locally deployed academic-paper repository and recommendation system. Users browse or import papers, extract and validate metadata, store PDFs and records, then search the repository or request ranked recommendations from a query or seed paper. The backend is FastAPI/SQLAlchemy over SQLite; the React frontend exposes repository, library, upload, recommendation, and evaluation workflows. Recommendation search combines configured TF-IDF, S-BERT, and metadata scores. Several frontend and backend subsystems are represented from README capabilities and the file tree; their internal wiring is not asserted where source excerpts do not show it.
Main entry points: api.py Paper API · models.py Paper records · search_service.py Recommendation search · App.tsx Web interface
Components 28
Web application
- Web interface
frontend/src/App.tsx - Repository workflows
frontend/src/pages/repository - Recommendation pages
frontend/src/pages/evaluation - Pipeline lab
frontend/src/pages/lab/Lab.tsx - HTTP client
frontend/src/api.ts
Paper repository
- Paper API
app/api.py - BibTeX extraction
app/services/bib_extraction.py - PDF extraction
app/services/extraction.py - Recommendation validation
app/services/validation.py - Subject classification
app/services/classification.py - Open-access PDF finder
app/services/pdf_finder.py - Repository search
app/services/fts_search.py - Repository queries
app/repositories/queries.py
Recommendation engine
- Text preparation
app/services/text_preparation.py - Recommendation search
app/services/recommendation/search_service.py - TF-IDF pipeline
app/services/recommendation/tfidf_pipeline.py - S-BERT pipeline
app/services/recommendation/sbert_pipeline.py - Metadata scoring
app/services/recommendation/metadata_pipeline.py - Pipeline weights
app/services/recommendation/pipeline_config.py - Vector index
app/services/recommendation/vector_index.py
Data and storage
- Paper records
app/models/models.py - SQLite database
app/database.py - Uploaded files
app/services/storage.py
Research features
- Evaluation runner
app/services/evaluation/runner.py - Research help chat
app/services/research_chat.py - Pixel pet and hunt
frontend/src/components/retro/PixelPet.tsx
Other components
- Researcher
- Online paper sources
Connections 38
- Researcher → Web interface: uses
- Web interface → Repository workflows: routes to
frontend/src/App.tsx - Web interface → Recommendation pages: routes to
frontend/src/App.tsx - Web interface → Pipeline lab: routes to
frontend/src/App.tsx - Repository workflows → HTTP client: requests data
frontend/src/pages/repository/Repository.tsx - Recommendation pages → HTTP client: requests data
- HTTP client → Paper API: HTTP requests
- Paper API → SQLite database: uses sessions
app/api.py - SQLite database → Paper records: maps records
app/database.py - Paper API → Repository queries: queries papers
app/api.py - Paper API → BibTeX extraction: imports citations
app/api.py - Paper API → PDF extraction: extracts metadata
app/api.py - Paper API → Recommendation validation: validates records
app/api.py - Paper API → Subject classification: classifies papers
app/api.py - Paper API → Open-access PDF finder: finds PDFs
app/api.py - Paper API → Repository search: searches records
app/api.py - Paper API → Uploaded files: stores uploads
app/api.py - PDF extraction → Paper records: populates metadata
- BibTeX extraction → Paper records: imports metadata
- Open-access PDF finder → Online paper sources: searches online
- Repository queries → Paper records: reads records
app/repositories/queries.py - Repository search → Paper records: searches records
- Paper API → Recommendation search: requests recommendations
app/api.py - Recommendation search → Text preparation: prepares query text
app/services/recommendation/search_service.py - Recommendation search → TF-IDF pipeline: scores lexical
app/services/recommendation/search_service.py - Recommendation search → S-BERT pipeline: scores semantic
app/services/recommendation/search_service.py - Recommendation search → Metadata scoring: scores metadata
app/services/recommendation/search_service.py - Recommendation search → Pipeline weights: loads weights
app/services/recommendation/search_service.py - Recommendation search → Paper records: selects candidates
app/services/recommendation/search_service.py - TF-IDF pipeline → Vector index: uses vectors
app/services/recommendation/tfidf_pipeline.py - S-BERT pipeline → Vector index: uses embeddings
app/services/recommendation/sbert_pipeline.py - TF-IDF pipeline → Paper records: reads corpus
app/services/recommendation/tfidf_pipeline.py - S-BERT pipeline → Paper records: reads corpus
app/services/recommendation/sbert_pipeline.py - Metadata scoring → Paper records: scores fields
app/services/recommendation/metadata_pipeline.py - Pipeline lab → HTTP client: sends recipe
- Researcher → Pixel pet and hunt: interacts
- Recommendation pages → Evaluation runner: evaluates results
- Paper API → Research help chat: answers help queries
app/api.py